CN-122016869-A - Online glass defect recognition system based on machine vision

CN122016869ACN 122016869 ACN122016869 ACN 122016869ACN-122016869-A

Abstract

The invention relates to the technical field of machine vision and discloses a glass defect online identification system based on machine vision, which comprises an image acquisition module, an image preprocessing module, a defect identification module, a result output module and an environment self-adaptation module, wherein an interference suppression unit in the environment self-adaptation module predicts and compensates environment illumination data acquired in real time by adopting a Kalman filtering algorithm, can suppress interference of environment light fluctuation, equipment vibration and light source attenuation on imaging, solves the problems of uneven image brightness and reflection spots caused by abnormal illumination, thereby ensuring the authenticity and stability of image features input into the defect identification module, reducing misjudgment and omission caused by image distortion, improving the detection precision and reliability of the system in a complex industrial environment, and continuously optimizing the global parameters of the system by adopting a reinforcement learning algorithm through a feedback unit, and improving the detection rate of micro scratches and pinholes.

Inventors

Zhou Yuzuo
ZHOU YUMENG
CHEN HAIXIA
YANG XIAOMEI
LIU JIANHUA
YU YUNYI
LIU PING
LIU HAI

Assignees

湖南省云迪钢化玻璃有限公司

Dates

Publication Date: 20260512
Application Date: 20260203

Claims (10)

1. A machine vision-based glass defect online identification system, comprising: The image acquisition module is used for acquiring image data of the glass coating surface by using the high-resolution camera unit, adjusting ambient light by using the illumination control unit to reduce interference, and outputting original image data by using the image transmission unit; the image preprocessing module is used for receiving the original image data, removing noise by a median filtering algorithm through a noise removing unit, improving image contrast by histogram equalization through an enhancement unit, and outputting preprocessed image data through a filtering unit; The defect identification module is used for receiving the preprocessed image data, performing defect detection by using a convolutional neural network through a deep learning model unit, identifying the defect type by using a classification unit based on a softmax function, and outputting a defect position coordinate through a positioning unit; the result output module is used for receiving the defect position coordinates and the defect types, calculating the defect size through the data fusion unit, generating a structured detection report through the report generating unit, and displaying a real-time result on a user interface through the visualization unit; And the environment self-adaptive module is used for receiving the real-time result and the environment sensor data, reducing the influence of environment fluctuation by adopting a self-adaptive filtering algorithm through the interference suppression unit, dynamically adjusting the detection threshold parameter by using the calibration unit, and optimizing the system performance parameter according to the historical data through the feedback unit.
2. The machine vision-based glass defect online identification system of claim 1, wherein the high-resolution camera unit of the image acquisition module comprises a CCD sensor and a telecentric optical lens, the illumination control unit comprises an adjustable LED array and a luminosity sensor, and the image transmission unit adopts a gigabit Ethernet protocol to transmit the original image data.
3. The machine vision-based glass defect online identification system of claim 1, wherein the denoising unit of the image preprocessing module performs noise suppression by adopting a wavelet transform algorithm, the enhancement unit improves image details by limiting a contrast adaptive histogram equalization algorithm, and the filtering unit comprises a Gaussian filter and a bilateral filter to output smooth preprocessed image data.
4. The machine vision-based glass defect online identification system of claim 1, wherein a deep learning model unit of the defect identification module performs defect region proposal based on FasterR-CNN architecture, the classification unit identifies defect types including scratches, bubbles and stains by using a multi-layer perceptron network, and the positioning unit outputs accurate coordinates of defects by means of bounding box regression.
5. The machine vision-based glass defect online identification system of claim 1, wherein the data fusion unit of the result output module integrates defect position, type and size information, the report generation unit outputs a detection report in an XML format, and the visualization unit realizes three-dimensional superposition display of defect results through an OpenGL library.
6. The machine vision-based glass defect online identification system of claim 1, wherein the interference suppression unit of the environment self-adaptation module compensates the change of the ambient light by adopting a Kalman filtering algorithm, the calibration unit dynamically updates the detection threshold value through an online learning mechanism, and the feedback unit optimizes the system parameters based on a reinforcement learning algorithm.
7. The machine vision-based glass defect online identification system of claim 1, wherein the defect identification module further comprises a feature extraction unit for extracting multi-scale defect features by using a residual network and enhancing the identification accuracy of the micro defects by an attention mechanism unit.
8. The machine vision-based glass defect online identification system of claim 1, wherein the result output module further comprises an alarm unit, and when the defect size is detected to exceed a preset threshold value, an alarm signal is output through an audible and visual alarm and is linked with a production line control system to realize automatic sorting.
9. The machine vision-based glass defect online identification system as set forth in claim 1, further comprising a real-time monitoring module, wherein the real-time monitoring module is used for receiving the original image data of the image acquisition module, temporarily storing the image stream through a frame buffer unit, detecting system delay through a delay analysis unit, and outputting a system real-time index through a performance evaluation unit.
10. The machine vision-based glass defect online identification system of claim 9, wherein the frame buffer unit of the real-time monitoring module adopts a ring buffer structure, the delay analysis unit calculates image processing delay through a time stamp, and the performance evaluation unit generates a system optimization suggestion based on throughput and accuracy.

Description

Online glass defect recognition system based on machine vision Technical Field The invention relates to the technical field of machine vision, in particular to a glass defect online identification system based on machine vision. Background Machine vision is a comprehensive technology integrating digital image processing, mechanical, control, lighting, optical and computer software and hardware technologies, and a machine vision system can realize contactless detection of products and rapidly acquire a large amount of information, and is easy to integrate with design information and processing control information when applied to industrial production, so that the machine vision technology is increasingly attractive as an important detection means and is widely applied to industrial detection. At present, as the glass coating production environment is complex and changeable, when on-line detection is carried out, the light projected on the glass surface by the illumination system is easily interfered by the fluctuation of ambient light, the vibration of equipment and the attenuation of a light source, so that the imaging brightness is uneven and anti-flare is generated, and when the illumination abnormality cannot be monitored and compensated in real time, the image characteristic distortion is directly caused, and further the misjudgment and the missed judgment of the defect are caused. Therefore, an online glass defect recognition system based on machine vision is proposed to solve the above problems. Disclosure of Invention Aiming at the defects of the prior art, the invention provides a glass defect online identification system based on machine vision, which solves the problems of uneven imaging brightness and generation of anti-light spots in the background art. In order to achieve the above purpose, the invention provides a glass defect online identification system based on machine vision, comprising: The image acquisition module is used for acquiring image data of the glass coating surface by using the high-resolution camera unit, adjusting ambient light by using the illumination control unit to reduce interference, and outputting original image data by using the image transmission unit; the image preprocessing module is used for receiving the original image data, removing noise by a median filtering algorithm through a noise removing unit, improving image contrast by histogram equalization through an enhancement unit, and outputting preprocessed image data through a filtering unit; The defect identification module is used for receiving the preprocessed image data, performing defect detection by using a convolutional neural network through a deep learning model unit, identifying the defect type by using a classification unit based on a softmax function, and outputting a defect position coordinate through a positioning unit; the result output module is used for receiving the defect position coordinates and the defect types, calculating the defect size through the data fusion unit, generating a structured detection report through the report generating unit, and displaying a real-time result on a user interface through the visualization unit; And the environment self-adaptive module is used for receiving the real-time result and the environment sensor data, reducing the influence of environment fluctuation by adopting a self-adaptive filtering algorithm through the interference suppression unit, dynamically adjusting the detection threshold parameter by using the calibration unit, and optimizing the system performance parameter according to the historical data through the feedback unit. Preferably, the high-resolution camera unit of the image acquisition module comprises a CCD sensor and a telecentric optical lens, the illumination control unit comprises an adjustable LED array and a luminosity sensor, and the image transmission unit adopts a gigabit Ethernet protocol to transmit the original image data. Preferably, the denoising unit of the image preprocessing module performs noise suppression by adopting a wavelet transformation algorithm, the enhancing unit improves image details by limiting a contrast self-adaptive histogram equalization algorithm, and the filtering unit comprises a Gaussian filter and a bilateral filter to output smooth preprocessed image data. Preferably, the deep learning model unit of the defect recognition module performs defect region proposal based on FasterR-CNN architecture, the classification unit uses a multi-layer perceptron network to recognize defect types including scratches, bubbles and stains, and the positioning unit outputs accurate coordinates of the defects through bounding box regression. Preferably, the data fusion unit of the result output module integrates defect position, type and size information, the report generation unit outputs a detection report in an XML format, and the visualization unit realizes three-dimensional superposition display of defect resul